Demis Hassabis, the cofounder and CEO of Google DeepMind, has called for the creation of a global artificial intelligence watchdog led by the United States. In a blog post titled 'A Framework for Frontier AI and the Dawning of a New Age,' Hassabis argued that as AI systems grow more sophisticated, the need for a coherent international regulatory framework becomes increasingly urgent. He envisions an organization modeled after existing regulators like the Financial Industry Regulatory Authority (FINRA), composed of independent experts and representatives from open-source communities, with the power to evaluate frontier models before they are deployed and to coordinate an industry-wide slowdown if the models are deemed too risky.
Hassabis’s proposal comes at a time when the AI industry is experiencing unprecedented growth and investment, but also facing mounting concerns about safety, bias, and the potential for misuse. The blog post, published in mid-July 2026, reflects a growing consensus among leading AI researchers that governance has not kept pace with technological capability. 'Artificial general intelligence is probably only a few short years away,' Hassabis wrote. 'When we look back on this time in the decades to come, I think we will realise we were standing in the foothills of the singularity - nothing less than the dawning of a new age for humanity.'
The proposed watchdog would be led by the United States, which Hassabis believes is the natural home for setting global standards 'given its economic and technical standing.' This endorsement of US leadership is significant, especially as other nations, particularly China and the European Union, have been developing their own approaches to AI regulation. The EU’s AI Act, for instance, is one of the first comprehensive legal frameworks for AI, but it focuses more on risk classification than on pre-release evaluation of frontier models. Hassabis’s idea positions the US as the primary architect of a new global oversight body, potentially influencing how AI governance evolves worldwide.
According to Axios, Hassabis has spent months quietly building support for his proposal. He has briefed the Trump administration, other AI labs such as OpenAI and Anthropic, and European officials. He told Axios that 'the noises I’ve been hearing from the Trump administration are very positive.' This suggests that the political climate in the US may be receptive to such an initiative, even as the country has been slower than Europe to enact binding AI regulations. Hassabis hopes to have the new organization up and running before the end of the year, a timeline that would require rapid coordination among governments, industry leaders, and civil society.
Hassabis is not new to the conversation about AI safety and governance. He was among the signatories of a statement last month calling for tougher protections against AI-aided bioweapons production. He also signed a similar statement in 2023 alongside hundreds of AI experts, warning that mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war. These efforts highlight his evolution from a researcher focused on building advanced AI to an advocate for responsible stewardship of the technology he helped create.
Born in London to a Greek Cypriot father and a Chinese Singaporean mother, Hassabis demonstrated exceptional cognitive abilities from a young age. He was a chess prodigy who achieved master status at age 13. After studying computer science at Cambridge University, he co-founded the video game company Elixir Studios before earning a PhD in cognitive neuroscience from University College London. He then co-founded DeepMind in 2010, which was acquired by Google in 2014. Under his leadership, DeepMind achieved landmark successes including AlphaGo’s victory over the world champion in the complex board game Go, AlphaFold’s breakthrough in protein structure prediction, and more recently, advances in AI systems that can navigate complex environments and generate human-like text.
In 2024, Hassabis was awarded the Nobel Prize in Chemistry jointly with John Jumper for their work on AlphaFold, an AI system that predicts protein structures. This recognition cemented his status as one of the most influential figures in both AI and science. The Nobel committee praised AlphaFold for solving a 50-year-old grand challenge in biology, demonstrating how AI can accelerate scientific discovery. However, Hassabis has consistently warned that the same technology can be used for harm if not properly regulated.
The proposal for a global AI watchdog is part of a broader push by industry leaders to establish norms before governments impose patchwork regulations. Hassabis’s blog post argued that 'the current state of AI governance is fragmented and reactive. We need a proactive, expert-led body that can move as fast as the technology itself.' The organization would have the authority to require that frontier models undergo safety evaluations similar to clinical trials for new drugs. If a model were found to pose unacceptable risks — for example, the ability to autonomously design novel pathogens or to manipulate critical infrastructure — the watchdog could order a temporary halt to deployment while fixes are developed.
Critics of such a proposal argue that it concentrates too much power in a single body, especially one led by the United States. They point out that many AI labs are based in the US, and a US-led regulator could stifle innovation or be captured by corporate interests. Others worry that other major AI powers, like China, would not submit to US oversight, potentially leading to a regulatory race to the bottom. Hassabis has responded to these concerns by emphasizing that the watchdog would be composed of independent experts and open-source representatives, not just government officials or corporate executives. He also noted that the body would work in concert with existing national regulators, not replace them.
The timing of Hassabis’s proposal is notable. Earlier in 2026, a group of top economists and tech titans, including Anthropic cofounder Jack Clark and former Google CEO Eric Schmidt, urged world leaders to take the looming economic impacts of AI seriously. Their statement highlighted the potential for AI to disrupt labor markets, widen inequality, and accelerate automation in unpredictable ways. Hassabis’s focus on safety complements these economic warnings, suggesting that the risks are multifaceted and require a coordinated response.
At the same time, AI capabilities continue to advance rapidly. Google DeepMind recently demonstrated a system capable of autonomously conducting chemistry experiments, raising questions about laboratory safety and dual-use applications. Other labs have unveiled models that can generate realistic video from text prompts, create persuasive political propaganda, and write computer code that could be used for cyberattacks. Each new capability brings the world closer to the kind of frontier models that Hassabis believes need oversight.
One of the key challenges for any global AI watchdog is defining what constitutes a 'frontier model.' Hassabis’s blog post suggests that threshold would be based on capabilities that could cause 'large-scale harm,' such as the ability to create weapons of mass destruction, disrupt critical national infrastructure, or destabilize global financial markets. The watchdog would maintain a list of such capabilities and update it regularly as technology evolves. Companies developing models that cross these thresholds would be required to submit them for evaluation before public release.
Another challenging issue is enforcement. If a company refuses to comply with a slowdown order, what happens? Hassabis proposes that the watchdog would work with national governments to impose penalties, including fines, export controls, or even criminal liability. However, such mechanisms would require international treaties or at least bilateral agreements, which are notoriously difficult to negotiate. The hope is that early adoption by the US and like-minded allies would create a de facto standard that other countries eventually join.
Despite these hurdles, the proposal has generated significant discussion in AI policy circles. Some see it as a more realistic alternative to the idea of a global moratorium on advanced AI development, which has been floated by some activists but lacks political support. Others appreciate that Hassabis is putting his weight behind a concrete policy proposal, rather than issuing vague warnings. The inclusion of open-source representatives in the watchdog’s composition is also seen as an attempt to address concerns about corporate control.
Hassabis’s role as a Nobel laureate and respected figure in the AI community gives his proposal added credibility. He is not merely a CEO advocating for his company’s interests; he is a scientist who has seen firsthand both the transformative potential and the risks of AI. His dual perspective — as both a developer and a regulator advocate — is rare in the tech industry, where most leaders either downplay risks or call for extreme caution. Hassabis navigates a middle path, arguing that innovation and safety are not incompatible if managed correctly.
The coming months will be crucial for the proposal’s fate. Hassabis will need to secure buy-in from the Trump administration, which has shown interest in AI but has also prioritized deregulation in many areas. He will also need to engage European officials, who are wary of US dominance in tech but may see value in a strong regulator. Meanwhile, other AI labs like OpenAI and Anthropic have their own governance proposals, and it remains to be seen whether they will support Hassabis’s plan or offer alternatives. The industry as a whole is watching closely, aware that the shape of future regulations will determine not only their business models but also the trajectory of AI’s impact on society.
Source: The Verge News